Predicting cardiovascular disease from real-time electrocardiographic monitoring: An adaptive machine learning approach on a cell phone

To date, cardiovascular disease (CVD) is the leading cause of global death. The Electrocardiogram (ECG) is the most widely adopted clinical tool that measures the electrical activities of the heart from the body surface. However, heart rhythm irregularities cannot always be detected on a standard resting ECG machine, since they may not occur during an individual's recording session. Common Holter-based portable solutions that record ECG for up to 24 to 48 hours lack the capability to provide real-time feedback. In this research, we seek to establish a cell phone-based real-time monitoring technology for CVD, capable of performing continuous on-line ECG processing, generating a personalized cardiac health summary report in layman's language, automatically detecting and classifying abnormal CVD conditions, all in real time. Specifically, we developed an adaptive artificial neural network (ANN)-based machine learning technique, combining both an individual's cardiac characteristics and information from clinical ECG databases, to train the cell phone to learn to adapt to its user's physiological conditions to achieve better ECG feature extraction and more accurate CVD classification on cell phones.